Since this forum is kinda dead anyway, I thought I would share some thoughts on the difference between us and the books and what that means for a bettor's strategy.
Maybe at this point it isn't a secret that I am not a fan of using statistics to develop betting strategies. I will admit that I've only taken one statistics class in my life, but if I took at least one useful thing from it, it would be this: statistics theories assume a static environment. It was developed to describe what we can know about an evenly distributed population from a random sample with a certain degree of confidence.
A great example of this is the exit polls of presidential elections versus the periodic phone polls that are taken in the months leading up to the election. The phone polls use a sample of "likely voters" (what if they don't vote?) and measure who they "think" they are going to vote for (what if they change their minds?). The exit polls are always right because that is statistics the way it was meant to be used -- to infer the distribution of a population from the distribution of a sample. You are asking people who definitely did vote exactly whom they voted for.
Statistics does not work with populations that are not evenly distributed. It also doesn't work with populations and samples whose sizes and components are in constant flux. Finally, there is no statistical theory that says that what happened in the past will happen in the future. This is a good anecdote to live by and certainly has some degree of truth, but how much and to what extent?
So the first thing we can throw out the window as bettors is all the crap about what makes a good sample size and what degree of confidence you can have based on a certain sample size, mean, standard deviation and all of that other theoretical stuff. Those theories were developed because mathematicians want to develop theories that can be applied universally to a static axiomatic system. Sports betting is neither static nor axiomatic. I believe it is best not to use statistics just because you feel you should, but to understand what these statistical measurements actually mean and develop your own equivalents for the precise situation that you need them.
Do the books use statistics? Sure, and one of the reasons is because their job is a hell of a lot easier than yours. Their job is to adjust the lines such that on average across all reasonably large samples of games given major variables, the difference between the results and the predictions fits within plus or minus the vig. They just need to make it so that if you bet a certain way on all home teams or all away teams or all favorites or all underdogs or all home teams coming off of a travel day, you will lose money in the long run. Once they have done that, they win at the game of statistics, because they know that once the cross-sections of games gets small enough (e.g. left-handed pitchers in Coors Field on a full moon with six days of rest), the sample size will be such that you cannot make a reliable prediction. It also doesn't hurt that their bankroll is infinite. So while you would be agonizing over a 10-game losing streak, your 10-game winning streak is a blip on the radar for the books. You cannot beat them at their own game.
Your job as a bettor is harder, but not impossible. You already know that the lines are a decent representation of most easy to procure variables -- you know what not to do, because the books already have a monopoly on that. Your job is to find a way to slice the data that produces consistent and reliable results despite that. It is possible to do so, but it requires a more diabolical approach. Your job is not to imitate the books -- your job is not to accurately represent sports games via regression modeling. That wouldn't be efficient. Your goal is to find a chink in the armor -- a system that thrives in the environment that the books create. That environment is well-suited to take the money of bettors chasing five- ten- and twenty-game "trends." If you can find a formula that makes as few assumptions as possible and remains profitable over thousands of games -- and preferably one that adapts over time to changing environments -- then you've got a winner.
Maybe at this point it isn't a secret that I am not a fan of using statistics to develop betting strategies. I will admit that I've only taken one statistics class in my life, but if I took at least one useful thing from it, it would be this: statistics theories assume a static environment. It was developed to describe what we can know about an evenly distributed population from a random sample with a certain degree of confidence.
A great example of this is the exit polls of presidential elections versus the periodic phone polls that are taken in the months leading up to the election. The phone polls use a sample of "likely voters" (what if they don't vote?) and measure who they "think" they are going to vote for (what if they change their minds?). The exit polls are always right because that is statistics the way it was meant to be used -- to infer the distribution of a population from the distribution of a sample. You are asking people who definitely did vote exactly whom they voted for.
Statistics does not work with populations that are not evenly distributed. It also doesn't work with populations and samples whose sizes and components are in constant flux. Finally, there is no statistical theory that says that what happened in the past will happen in the future. This is a good anecdote to live by and certainly has some degree of truth, but how much and to what extent?
So the first thing we can throw out the window as bettors is all the crap about what makes a good sample size and what degree of confidence you can have based on a certain sample size, mean, standard deviation and all of that other theoretical stuff. Those theories were developed because mathematicians want to develop theories that can be applied universally to a static axiomatic system. Sports betting is neither static nor axiomatic. I believe it is best not to use statistics just because you feel you should, but to understand what these statistical measurements actually mean and develop your own equivalents for the precise situation that you need them.
Do the books use statistics? Sure, and one of the reasons is because their job is a hell of a lot easier than yours. Their job is to adjust the lines such that on average across all reasonably large samples of games given major variables, the difference between the results and the predictions fits within plus or minus the vig. They just need to make it so that if you bet a certain way on all home teams or all away teams or all favorites or all underdogs or all home teams coming off of a travel day, you will lose money in the long run. Once they have done that, they win at the game of statistics, because they know that once the cross-sections of games gets small enough (e.g. left-handed pitchers in Coors Field on a full moon with six days of rest), the sample size will be such that you cannot make a reliable prediction. It also doesn't hurt that their bankroll is infinite. So while you would be agonizing over a 10-game losing streak, your 10-game winning streak is a blip on the radar for the books. You cannot beat them at their own game.
Your job as a bettor is harder, but not impossible. You already know that the lines are a decent representation of most easy to procure variables -- you know what not to do, because the books already have a monopoly on that. Your job is to find a way to slice the data that produces consistent and reliable results despite that. It is possible to do so, but it requires a more diabolical approach. Your job is not to imitate the books -- your job is not to accurately represent sports games via regression modeling. That wouldn't be efficient. Your goal is to find a chink in the armor -- a system that thrives in the environment that the books create. That environment is well-suited to take the money of bettors chasing five- ten- and twenty-game "trends." If you can find a formula that makes as few assumptions as possible and remains profitable over thousands of games -- and preferably one that adapts over time to changing environments -- then you've got a winner.
